Add API for GrContext to recommend rendertarget sample count

Add API for GrContext to recommend rendertarget sample count

Add GrContext::getRecommendedSampleCount method that can be
used to determine which GPU backend and what exact sample count
is recommendeded for a particular render target at particular
dpi.

Make this initially recommend 4xMSAA for contexts which have
NVPR enabled if dpi is 250 or more, 16 if dpi is less than 250
and no MSAA for others.

BUG=chromium:347962
Committed: http://code.google.com/p/skia/source/detail?r=13717

[Kimmo Kinnunen, 2014-03-06 07:15:31]

How about this?

The Chromium call site patch is:
https://chromiumcodereview.appspot.com/184573002/

[bsalomon, 2014-03-06 14:02:43]

the interface looks good

https://codereview.chromium.org/188403003/diff/1/src/gpu/GrContext.cpp
File src/gpu/GrContext.cpp (right):

https://codereview.chromium.org/188403003/diff/1/src/gpu/GrContext.cpp#newcod...
src/gpu/GrContext.cpp:1709: return 4;
Should probably also check that 4 >= caps()->maxSampleCount().

I'm thinking that we may want to check the DPI and go to 16 samples on a typical
"desktop" display. Maybe if we're below 250 dpi? I'm sure what the right
threshold is but I think 4 samples will look bad on a standard monitor. (Maybe
in the future someone can do some experimentation and fine tune it and perhaps
figure out if it makes sense to have an 8 sample middle ground as well.)

[Kimmo Kinnunen, 2014-03-07 13:00:22]

new version

[bsalomon, 2014-03-07 18:35:36]

https://codereview.chromium.org/188403003/diff/40001/src/gpu/GrContext.cpp
File src/gpu/GrContext.cpp (right):

https://codereview.chromium.org/188403003/diff/40001/src/gpu/GrContext.cpp#ne...
src/gpu/GrContext.cpp:1718: return GrMin(fGpu->caps()->maxSampleCount(),
chosenSampleCount);
maybe:

return chosenSampleCount < fGpu->caps()->maxSampleCount() ? chosenSampleCount :
0

?

I think we should not recommend MSAA if we can't meet some min AA quality bar.

Otherwise lgtm (feel free to update and check CQ if you're ok with the
suggestion).

[Kimmo Kinnunen, 2014-03-10 07:24:33]

On 2014/03/07 18:35:36, bsalomon wrote:
> maybe:
> 
> return chosenSampleCount < fGpu->caps()->maxSampleCount() ? chosenSampleCount
:
> 0
> 
> ?
> 
> I think we should not recommend MSAA if we can't meet some min AA quality bar.

Did you mean "less than max sample count" or "less or equal to max sample count"
?

I assumed the latter. If that was incorrect, I should fix it in a later commit.

[Kimmo Kinnunen, 2014-03-10 07:25:03]

The CQ bit was checked by kkinnunen@nvidia.com

[commit-bot: I haz the power, 2014-03-10 07:25:12]

CQ is trying da patch. Follow status at
https://skia-tree-status.appspot.com/cq/kkinnunen@nvidia.com/188403003/60001

[commit-bot: I haz the power, 2014-03-10 07:40:07]

Change committed as 13717

[bsalomon, 2014-03-10 14:00:02]

On 2014/03/10 07:24:33, kkinnunen wrote:
> On 2014/03/07 18:35:36, bsalomon wrote:
> > maybe:
> > 
> > return chosenSampleCount < fGpu->caps()->maxSampleCount() ?
chosenSampleCount
> :
> > 0
> > 
> > ?
> > 
> > I think we should not recommend MSAA if we can't meet some min AA quality
bar.
> 
> Did you mean "less than max sample count" or "less or equal to max sample
count"
> ?
> 
> I assumed the latter. If that was incorrect, I should fix it in a later
commit.

Yes, you're right.